PG Bug reporting form <noreply(at)postgresql(dot)org> writes:
> SELECT position('foo'::citext IN 'Foobar'::citext) =
> strpos('Foobar'::citext, 'foo'::citext) as "positionEqualsStrpos";
> Citext does not create an overload for position() supporting citext
> parameters, therefore position(a in b) always runs case-sensitive.
Well, the citext documentation specifies which functions have
case-insensitive mappings. strpos() is listed, position() is not,
so I'd say it's acting precisely as documented.
We could consider adding position() of course, but there's a
backwards-compatibility issue --- at this point, people may have
queries that depend on the current behavior.
regards, tom lane